Effects of Operation Prediction Sharing for Collaborative Avatar Robot

A collaborative avatar robot, in which multiple users cooperate with a single-robot avatar, improves operability by leveraging the judgment and skills of both operators. However, collaboration among users is essential for a smooth operation. In this paper, we propose a system that provides a visual display of operational predictions to facilitate mutual understanding of actions among users in a single robot arm as a collaborative avatar robot. The robotic arm was controlled by two operators, while the system predicts operations 0.3 seconds ahead using machine learning based on one operator's actions. These predictions were visually presented to the other operator. Pick-and-place experiments were conducted, where two operators were assigned roles as leader and follower, with only the leader knowing the correct block to manipulate. The task performance and operability were evaluated with and without prediction system. Additionally, since operability and workability are affected by the control ratio of the two operators, three control ratio conditions were evaluated for comparison purposes. The results suggest that the followers' subjective ratings in the sense of agency, smoothness, and workload significantly improved and that the operation effort of the leader significantly decreased while maintaining the sense of agency, reducing timing discrepancies between the operators.